Tuesday, 20 April 2010

my new friend, node number 5

The Dear Reader of this blog probably does not know that before I started to go down the route of academia, I used to run a research team. This had some perks. To start with the obvious, my research team had a budget, a marked difference from the current state of affairs. As we were an 'emerging market research team' we covered a lot of countries. As a consequence, I got to explore a lot of places, and more importantly, when I felt that something important was about to happen in a country, I simply hopped on the plane, went there and checked. This was very-very interesting.

But there were also other perks. Chiefly, when I had a new idea, I just asked someone to check it. It was the simplest thing in the world. I wondered what the inflation patterns during structural change were? Did there happen to be a uniform framework applicable for all? Are the institutional tools always different during the different phases? How about late comer countries? And so on. I would convene a small meeting, dish out the jobs, and wait for the results. It was wonderful.

Now, it is quite different. On one hand, I love that I do not have to travel. But, on the other hand, when I have a new idea, most of which usually turns out silly (the rate has not improved), I have to do all the checking myself. This takes a long time... The accumulated still-to-be investigated silly idea pile has grown substantially the past two years. So, not everything is 'cream to the bottom' as the Hungarian saying goes.

But, at least, my computer processors do not have personalities... Processor number five does not turn out to be hopelessly in love with processor number one. Processor number one does not want to switch to yoga instead of running my models, and processor number eight does not decide to leave the computer for a better ventilated slot in the middle of the calculation peak. Most importantly, my processors do exactly what I ask them to do. There is no squabbling, no turf fights, no 'well-meant simplifications'. Just the job. The way I wanted it. What a delightful life, really.

Until the world changes for the darker side again...

A few days ago, I woke up thinking about my calculations. This is nothing unusual, most mornings start with a semi-coherent -- and, I am being regularly reminded, far-far too long -- ramble about the latest models. Which recently tend to be about social structures emerging via information synchronisation Dunbar-style. Virtual societies live on my computer, and virtual friendship dynamics play out within them.

On that morning I also woke up with the simulations in mind. But, this was an exception. I was not thinking about a model, or a calculation problem. Instead, I was thinking about a node. One node. Node number 5, in particular.

In the model, in which node number 5 acquired prominence, there are 20 nodes, and they are all useful for the group (to be more precise, they have some probability of receiving the information about the direction of the hunt). The model is set up in a way that their usefulness (the probability of being the Chosen One receiving the vital hunting information) varies: the higher your node index, the higher the usefulness is. That is, node number 1 is the bottom of the social meritocracy, node number 20 is the top. You can see that node number 5 is not very high up.

If you introduce some group dynamics in which the nodes tend to choose their limited number of friends based on the other's usefulness, you are likely to see a stratified social structure. The top guys are friends with the top guys, the bottom guys are friends with the bottom guys. But there is also some hysteresis in the structure. For these nodes assess each other's merit based on the quality of hunting information they get from them. So if you are a lowly number 5, but you happen to be friends with nodes number 18, 19 and 20, then you will get excellent quality information all the time, and thus others will think that you are up that league. But, you -- I know, I coded you -- are really not. You are just node number 5, who has some childhood friends in high places...

In most cases, the group stratifies according to the true underlying merit fairly fast. But this case was different. Node number 5 happened to start the simulation rounds with so many top friends, that every other node got convinced about his great abilities, gave him high status, and he just stayed near the top forever. Two days of simulations passed, and this guy was still there!

So, on that morning, thinking about node number 5, I found myself thinking about his relationships with his partners, that he actually dropped some low level friends, and managed to acquire new top friends. And the fact that he was widely accepted by this mini-society as useful way beyond his true merits. I was wondering what happened to him overnight, and what I would find on my screen once I get to my study. I told my wife about him, and passingly mentioned him to my young children (who surely must think that Anything Is Possible on my computer...) In other words, I was properly gossiping about him.

So here is the question: is node number 5 my friend now? By the Dunbarian definition of a 'friend', he surely is. I got engaged in his social life, I gossiped about him, speculated, passed on rumours, and clearly assigned intentions to a by-definition intentionless thingy on my computer. Worst of all I was emotionally engaged with his fate, proved by the fact that despite thinking that he was a rascal, I felt sad when I saw him fall from grace, which he ultimately was doomed to do. If a soap opera figure can be a virtual friend, node number 5 became certainly at least as good a friend for me.

But then, he does not know about me. It is only me who gossips, speculates, gets emotionally involved. This really is a very one-sided friendship. He is perhaps not a friend, but more a celebrity. One that, now that you have read this, has, maybe, one extra fan: you. But, he is not your friend, he does not care about you. And anyway, he is merely a dot on some strange blogger's screen.

Don't forget this, when you mention him passingly to your acquaintances...

Monday, 29 March 2010

is altruism really always self-interested?


Economics is a one-idea discipline: incentives work. All the rest is engineering. Perhaps the arrogance of economists come from the fact that this idea sometimes seems exclusive to their discipline. Anyway, incentives work. Or to put it perhaps more controversially (and truthfully) incentives are the only thing that work... If so, then behaviour patterns that are the product of evolution but which appear altruistic must be either (a) a mistake, or (b) ultimately self-interested.

Society-thinkers have long been perplexed by the act of 'goodness'. Social engineers seize on the human propensity to be good to others to build institutions. But theorists keep scratching their heads. There is something fishy here. People seem to be good - sometimes, at least - even when it is not in their interest.

The first explanation posited a biological concept of altruism, based on the observation that animals seemed to be 'good' primarily towards their kin. Although altruistic (or more precisely: seemingly-altruistic) behaviour is widespread in the biological world, it all but disappears once genetic relationships are accounted for. It is the selfish gene that drives kin-selection. (For people of my generation these concepts are obvious. Thus it was very educating to read Lee Alan Dugatkin's book on the history of the biological concept [1], just to be reminded that concepts that we take for granted have not been 'the truth' for long...)

Yet, there is something else: some other kind of altruistic behaviour, widespread especially among humans, which cannot be explained by kin-selection. Being good to somebody else with whom one has a very distant genetic relationship, without any reason whatsoever, just should not happen under the biological theories. Social altruism was thus left unexplained.

One group of theorists suggest that the explanation lies in group selection. The trouble is that group selection is very unlikely to have happened due to a set of mathematical properties. (Although I have been raised in the school that allows for group selection, sadly and literally, 'after much consideration', I must turn against the tradition in which I was brought up and leave this path to the philosophers...)

So we are left with a truly social phenomenon.

Enter the economists... Game theory provides a brilliant set of models about reciprocity and reputation. (For an overview and some models, see [2] and [3].) First, you are likely to be 'altruistic' to others if you expect that they will return the favour in the future.  A two-person insurance pool. Second, you might see it as advantageous to acquire a reputation of being altruistic, if that makes people be more altruistic towards you when you are in need: reputation facilitates indirect reciprocity. This allows a much wider insurance pool. (The cynic must admit though that a good way to be seen as an altruist is probably to be one.)

These biological and societal explanations span a field in which altruism acquires a particular set of attributes.

One. There is nothing altruistic in altruism. Altruism is either not altruistic at all (kin selection), or it is kind-of altruistic, but with expectations about eventually receiving something in return (reciprocity).  Or it is about pretending to be altruistic in a very self-interested way (reputation). The idea of economics, that incentives work, prevails...

Two. Structure is conspicuously missing from the models, especially when it comes to human groups. Unless I have missed that part of the literature, it seems that the models out there discuss altruistic behaviour regarding 'others in general', rather than others as in 'other people of a particular group of a particular structure'. It would make a lot of sense to try to model the evolution of social altruism as an element of group dynamics.

Group-level modelling would allow us to understand much more of the variation in altruistic behaviour that can be observed. For instance, why did some people during World War Two risk their lives to save Jews that they had never met, even in a foreign country, while others did not do anything. Was Wallenberg really a selfless altruist, or was he rather a member of a particular society with a framework of norms within which his altruistic behaviour could be interpreted. To put it as controversially as possible: was Wallenberg accumulating his reputation capital? The space for discussion of evolving reference groups in the era of societal globalisation would be infinitely large...


[1] Dugatkin, L.A., (2006), The Altruism Equation (Princeton). Link: The Altruism Equation: Seven Scientists Search for the Origins of Goodness

[2] Nowak, M.A. and K. Sigmund, (2005), Evolution of indirect reciprocity, Nature. Link: Evolution of indirect reciprocity

[3] Elreath, R., and R. Boyd, (2007), Mathematical Models of Social Evolution, (Chicago). Link: Mathematical Models of Social Evolution: A Guide for the Perplexed